Research map / Label-efficient and robust learning / OOD and anomaly detection
OpenOOD: Benchmarking Generalized Out-of-Distribution Detection
Jingkang Yang, Pengyun Wang, Dejian Zou, Zitang Zhou, Kunyuan Ding, WENXUAN PENG, Haoqi Wang, Guangyao Chen, Bo Li, Yiyou Sun, Xuefeng Du, Kaiyang Zhou and 4 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use simple data-augmentation methods such as PixMix instead of other evaluated method types | OOD detection · CIFAR-10 OOD benchmark | Near-OOD AUROC: 93.1% | Improved |
| Use extra outlier data with UDG rather than the strongest data-free method, KNN | OOD detection · CIFAR-10 near-OOD | AUROC: 91.9 vs 90.5 | Improved |
| Use extra outlier data with UDG rather than KNN | OOD detection · CIFAR-100 OOD benchmark | AUROC: 75.8 vs 79.9 | Hurt |
| Use a post-hoc OOD method instead of a method requiring training | OOD detection · generalized OOD detection benchmarks | AUROC | Mixed |
| Apply data augmentation to improve uncertainty estimation | OOD detection · CIFAR-10 and CIFAR-100 OOD benchmarks | AUROC: 93.1%; 85.5% | Improved |
| Use a detector evaluated on ImageNet instead of CIFAR benchmarks | OOD detection · ImageNet, CIFAR-10, and CIFAR-100 OOD benchmarks | AUROC | Mixed |
| Use recent KNN rather than classic MDS | post-hoc OOD detection · CIFAR-10 and CIFAR-100 OOD benchmarks | AUROC: 90.5 vs 66.5; 79.9 vs 51.4 | Improved |
| Use anomaly-detection methods CutPaste or DRÆM for far-OOD detection | OOD detection · CIFAR-100 far-OOD | AUROC: 83.3; 85.4 | Improved |
| Use OSR methods for near-OOD detection | OOD detection · near-OOD benchmarks | AUROC | Improved |
| Use external outlier data for OOD training | OOD detection · OOD detection benchmarks | AUROC | No clear effect |
Closest papers by meaning
- MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities (NeurIPS 2024 · similarity 0.90)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.90)
- Rethinking Out-of-Distribution Detection on Imbalanced Data Distribution (NeurIPS 2024 · similarity 0.89)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.89)
- What If the Input is Expanded in OOD Detection? (NeurIPS 2024 · similarity 0.89)
- Out-Of-Distribution Detection with Diversification (Provably) (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
